Neural Radiance Fields (NeRF) is a deep learning-based method that can reconstruct 3D scenes and synthesize new viewpoints from a set of input images. NeRF has enabled significant advancements in areas such as 3D scene understanding, novel view synthesis, human body reconstruction, and robotics.
GaussianAvatars is a method that creates photorealistic and fully controllable head avatars from multi-view video recordings. It represents the head as a dynamic 3D radiance field using 3D Gaussian splats rigged to a parametric face model, enabling precise animation of the avatar's expression, pose, and viewpoint.